... assign generation probabilities to NULLs. Re-garding the training corpus that will be described in Section 3, the NULL rates of the nominative,accusative, and dative cases were 16.7%, 59.9%,and ... claim and add two different kinds of features. One is the feature thatindicates whether a candidate has been used as anargument of predicates in the preceding sentences(‘Used’ features). However, ... articles. We dividedthem into training, development, and test sets asshown in Table 2.Argument Identification Models: Maximumentropy models were trained using the training set. In these experiments,...